AI Literacy Under Article 4 of the EU AI Act: What Your Organization Should Do Now
As of November 2026 · Not legal advice
Since August 2, 2026, authorities in the EU have been checking whether companies train their employees to work with AI. The basis is Article 4 of the EU AI Act: the AI literacy obligation, often called the AI training requirement. It applies to organizations in the EU whose people use AI at work, and it can reach companies outside the EU too, for example when they offer AI systems in the EU or when the output of their AI is used there.
It is worded more openly than many think: The law requires that training takes place, but not how good it has to be. Checking a box won't get you anywhere with AI. In this guide, we show you what the AI Act means for companies doing business in the EU, and how to meet the obligation in a way that benefits your employees and your company.
- Since February 2, 2025
- Companies must take measures on the AI literacy of their employees.
- Since August 2, 2026
- Authorities check compliance. Each EU country has its own market surveillance authorities.
- No specific fine
- Article 4 has none. But authorities can order measures in individual cases.
- No certificate
- The European Commission makes it clear: You don't need a certificate for Article 4.
When Did the AI Literacy Requirement Take Effect, and Who Does It Apply To?
Article 4 has applied since February 2, 2025, and authorities have been checking compliance since August 2, 2026. It applies to every organization in the EU whose employees use AI tools at work, and it can also reach companies outside the EU that do business there.
- February 2, 2025 Article 4 applies From this day on, companies must take measures on the AI literacy of their employees.
- July 27, 2026 New wording Regulation (EU) 2026/1744 changes the wording of Article 4. Companies now have to support the development of AI literacy instead of ensuring it, and no one has to guarantee a specific level anymore. More on this in the law.
- August 2, 2026 Authorities check compliance The national market surveillance authorities of each EU country are responsible.
Who it applies to
Organizations in the EU whose staff use AI at work, and companies outside the EU whose AI output is used there. Using ChatGPT or Copilot makes you a deployer, too. Personal use is exempt.
What happens if you don't comply
Article 4 has no specific fine. But authorities can order measures in individual cases, and these must be proportionate. You'll find more on this in the FAQ below.
Our advice: Don't try to just check a box. The EU kept Article 4 so vague that it's unclear whether that is even possible. You can read why in the sections on the law, on your leeway and on the standard. Unfortunately, many providers still use the fear of penalties to push for quick contracts. Fear is a poor advisor. Instead, design an enablement program that truly moves your employees and your company forward. Then the Article 4 obligation is met almost along the way. A good first step is an overview of which AI tools are actually used in your company, including the ones that were never officially introduced.
What Does Article 4 of the EU AI Act Require?
Providers and deployers of AI systems must take measures to support the development of AI literacy of their staff. Since the amendment of July 27, 2026, no one has to guarantee a specific level anymore.
“Providers and deployers of AI systems shall take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf, taking into account their technical knowledge, experience, education and training and the context the AI systems are to be used in, and considering the persons or groups of persons on whom the AI systems are to be used. This obligation does not require providers or deployers to guarantee any specific level of AI literacy of any individual.”
Art. 4(1) of the AI Act, Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744. Source: AI Act Service Desk of the European Commission
“Providers and deployers of AI systems shall take measures toensure, to their best extent, a sufficient level of AI literacyof their staff and other persons dealing with the operation and use of AI systems on their behalf …”
Art. 4 of the AI Act, Regulation (EU) 2024/1689, original version (OJ L, July 12, 2024). Source: EUR-Lex
Support, not ensure
The old wording could be read as making companies answerable for a result. Now it is explicitly about the effort.
No guaranteed level
No company has to guarantee that an individual reaches a specific level. That is now written into the law.
Support for SMEs
The EU and its member states are to help companies, especially SMEs. An EU body is to recommend common goals.
What this means for you: The obligation and the standard have stayed the same. The measures have to fit what a person knows and how they use AI. One course for everyone is therefore unlikely to be enough. You need different learning journeys for different roles. And short courses on compliance and governance won't do the job. It's not just about using AI safely, but about using it productively. That takes a broader view that includes leadership.
What Does Article 4 Not Require?
Article 4 says that companies must take care of sufficient AI literacy, but not how. According to the European Commission, there are no fixed requirements for this:
- No mandatory curriculum
- No minimum number of hours
- No mandatory format, whether in person, online or self-study
- No certificate. The Commission explicitly says that none is needed.
- No guaranteed level of competence for individuals
Source: AI Literacy – Questions & Answers by the European Commission
There is no guarantee. Even if you adopt the Commission's examples, you are not automatically compliant, according to the Commission. No one can guarantee legal certainty today, including us.
For L&D, that is an opportunity. If there is no safe checkbox, a higher standard pays off: training designed to the best of your knowledge that truly empowers your employees and delivers real business value. That also makes it easier to justify the budget to senior management. What appropriate means is shown in the next section.
What Does “Appropriate” Mean for AI Literacy?
Appropriate measures fit what a person knows and how they use AI. One course for everyone is therefore unlikely to be enough.
- The person
- Technical knowledge, experience and education
- The use
- The context in which the AI systems are used
- The people affected
- The people or groups on whom AI is used
According to the law, AI literacy also covers opportunities and risks. So it is about productive use, not just safe use.
Learning journeys by role
The Commission confirms that different levels make sense depending on prior knowledge and role. A manager who decides how AI is used needs something different from someone who writes texts with it every day.
Who received what?
Documentation is not mandatory. But the Commission recommends keeping internal records of which trainings and measures took place.
Our view: An attendance list shows that someone was there. It doesn't show whether the training fit that person and their work. The law doesn't say this word for word. But if an authority looks into an individual case, it will ask exactly that. A checkbox doesn't answer that question. An enablement program with learning journeys by role, as we recommend in our advice, does. And it moves your company forward with AI.
How Can You Tell Whether AI Training Works?
We measure training against Donald Kirkpatrick's model, with four levels at which it can have an effect. Article 4 doesn't even reach the first one.
Article 4 requires that a measure exists, but not that people find it useful. We call that level 0. With our customers, we want to achieve something on all four levels above it.
-
Level 0
The obligation
There is a measure, and it is documented. Article 4 asks for nothing more.
-
Level 1
Reaction
Participants experience the training as real added value for their work. Over the past 15 years, our Net Promoter Score has been 97. After every course, we collect feedback and review it with you.
-
Level 2
Learning
Participants learn something. Henrik, our AI coach, makes sure of that. At the start, he finds out in a conversation where each person stands. During and after the course, he is their sparring partner and responds to exactly that level. The assessment, also a conversation, shows what has changed.
-
Level 3
Behavior
Participants work differently than before. That's why they practice on their own tasks in the Playground. After the course, Henrik keeps following up and helps them bring what they learned into their daily work. In our view, there is no more effective way to help people adopt new behavior.
-
Level 4
Results
Your employees actually achieve more, each person in their own field. Henrik learns what has improved in his sparring with participants. We prepare this for you so you can see what the training brings your company.
The EU legislator, too, sees the obligation as only a starting point. The reasoning behind the amendment states:
AI literacy “should be a strategic priority, regardless of regulatory obligations and potential sanctions”.
Source: McKinsey, The State of AI, survey from August 2026, 1,719 respondents
Why this matters to us: McKinsey identified eleven practices that set these companies apart from the rest. But new workflows only emerge if the people working in them can use AI productively. That is exactly where good training comes in.
What we promise and what we don't: With Campus, you can show that you have taken suitable measures for different roles. But no course makes you compliant, ours included. What matters to us is that the training pays off for you.
What Does It Take for AI Training to Pay Off?
A course alone isn't enough. For training to pay off for your company, level 4 in the Kirkpatrick model, three things have to come together: Access, Enablement and Redesign.
Access
Your employees have the right AI tools, set up securely. Which ones is your decision. Our content works with any tool, even if you switch later. You can start with enablement before that, in the Playground.
Enablement
Basic understanding isn't enough: your employees need practice on their own tasks. In Campus, they learn at their own pace, with Henrik by their side into daily work. In the Playground, they try it out right away.
Redesign
The biggest impact comes when workflows are rethought, not when old ones run a little faster. Leaders provide the impetus. In AI for Leaders, they learn what AI means for their area and start the redesign in their teams.
What sets us apart: Many trainings stop at enablement. We combine it with redesign: Leaders learn to rethink workflows in AI for Leaders, and their teams get the tools in Campus to put that into practice. Only when both come together does level 4 become achievable.
How We Help You Build AI Literacy
Campus with Henrik, the Playground and our Live Trainings AI for Leaders and AI for Practitioners are available right away. Since November 2026, Campus also includes the first course from our AI literacy learning path, with more topics to follow.
Available now
Campus with Henrik
Assessment at the start, sparring during and after the courses, an assessment conversation at the end and a report for you.
Playground
Our protected practice environment in Campus. Here, participants build their own agents and small tools without coding.
AI for Leaders
For leaders: what AI means for leadership. Also available as e-learning in Campus from November 2026.
AI for Practitioners
For everyone who uses AI in their daily work. Coming soon as e-learning in Campus as well.
Custom Journeys
We already put together custom learning journeys for companies today, as a mix of Live Trainings and e-learning. Several customers have completed them, with very good feedback.
Which Topics Does Our AI Literacy Learning Path Cover?
Our learning path is a roadmap of sixteen topics in four areas: basics, practice, agents and automation, and topics for specific roles. We aim to cover all of them by the end of 2027. Whether a topic becomes its own course or part of a larger one, we decide during development. The roadmap itself may also still change along the way.
Basics
For everyone, no prior knowledge
Practice
For your daily work
Beyond Chat
Agents and automation
By Role
For special responsibility
Learning paths by role
| Basics | Practice | Beyond Chat | By Role | AI for Leaders | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | ||
| All employees | |||||||||||||||||
| Most employees | |||||||||||||||||
| Experienced users | |||||||||||||||||
| Leaders | |||||||||||||||||
| Oversight in high-risk areas | |||||||||||||||||
Our tip: With any training, pay attention to how personal it is. The better it fits a person's role and work, the more likely new behavior is to emerge, and this context of use is exactly what Article 4 names, too. With us, participants already state their function today, for example sales, HR, finance or engineering, and Henrik adapts examples and exercises to their work. By Q3 2027 at the latest, we'll go further: fully individual learning journeys for every person.
When Are the Next Topics Coming?
The basics are set to be in place by the end of 2026. In 2027, we plan a new module every quarter: practice, agents and automation, individual learning journeys and finally the topics by role. The roadmap may still change as development progresses.
- Already available Live Trainings, Custom Journeys, Campus AI for Leaders and AI for Practitioners as Live Trainings, Custom Journeys for companies, plus Campus with Henrik and the Playground.
- November 2, 2026 AI Foundations: Seeing AI Clearly The first course in the basics, in Campus. Available in English, more languages are in production and planning.
- November 2026 AI for Leaders as e-learning The Live Training will then also be available in Campus.
- By the end of 2026 Basics Topics 2 to 4, for all employees.
- Q1 2027 Practice Topics 5 to 9, the skills for your daily work.
- Q2 2027 Beyond Chat Topics 10 to 13, agents and automation.
- Q3 2027 Individual learning journeys Fully individual learning journeys for every person, matching their role, level and tasks.
- Q4 2027 Topics by role and for high-risk AI Topics 14 to 16, including “Overseeing AI That Makes Decisions About People”.
Build AI Agents Without Coding
In the Playground of the Agile Academy Campus, your employees try out what they've learned right away. They build their own AI agents and small apps for their own tasks and practice prompting, without coding and in a protected environment. Today, it offers the modules Agents, Coding and Chat, and more tools will follow with the next courses.
- Agents: build your own AI agents
- Coding: build small apps simply by describing what you need
- Chat: practice prompting and compare answers
Simplified prototype.
Try it out: Click the cards and buttons in the window.
Enablement before rollout. Because the Playground is a separate, secure environment, you don't need to have rolled out AI in your company yet to start training. Your employees are already learning and practicing while you decide on tools and approvals.
What sets us apart. A practice environment that grows with the courses is a fixed part of our training. Most providers don't offer that.
How Do You Document AI Literacy Under Article 4?
Documentation is not mandatory. But the Commission recommends keeping internal records of who received which training. Campus takes care of that for you and shows at the same time whether the training works.
Where people stand
At the start, Henrik has a short assessment conversation with each person. This results in their starting level and their personal learning path through the courses.
Who's progressing
You see who is making progress, who is stuck and where the group stands. Surveys and submissions can be exported, and anonymized if your employee representatives ask for it.
What people can do
The assessment is a conversation with Henrik. It evaluates whether someone can apply what they learned. You don't need a certificate for Article 4 (see leeway), but ours shows real skills.
What stays private
Conversations with Henrik are private: neither managers nor admins can see them. That protects your employees and makes buy-in from them and their representatives easier.
Tailored to your requirements: Every company documents differently. That's why we work out with every customer what you specifically need internally, for example for HR, legal or employee representatives, and set up the admin area accordingly.
How Much Does AI Training With Campus Cost?
With Campus, your employees get access to all available online courses and all courses added during your subscription, including the option to get certified, the AI coach Henrik, the Playground and the community.
We are raising the price on January 1, 2027. If you book before then, you still get the current price.
Until December 31, 2026
Applies for the entire term of the subscription.
From January 1, 2027
For bookings from January 1, 2027.
- All online courses, including new ones during your subscription
- AI coach Henrik
- Playground
- Option to get certified
- Community
Your price advantage when you book by December 31, 2026: You pay €499 instead of €999 per person per year, not just in the first year, but for as long as your subscription runs. An example: With 50 employees, you save €25,000, every year.
What Else Applies to High-Risk AI (Article 26)?
If you deploy high-risk AI, for example in recruiting, Article 26(2) requires you to assign human oversight to people with the necessary competence, training and authority. Unlike Article 4, violations can lead to fines here.
- Where it applies
- For example recruiting, credit checks, and life and health insurance
- What applies
- Oversight by people who are competent, trained and authorized for it
- From when
- For most of these systems, the obligation applies from December 2, 2027
- What's at stake
- Fines, unlike Article 4, which doesn't provide for any fines of its own
As a deployer, you have to assess yourself whether a system falls under this.
How Seriously Authorities Take Real Human Oversight
Source: Autoriteit Persoonsgegevens (Dutch Data Protection Authority), press release of August 21, 2026
Our advice. Clarify early whether any of your systems count as high-risk AI. The deadline is December 2, 2027, and the people taking on oversight need lead time.
How we support you. We already offer implementation consulting together with experts from our network. Topics from our learning path lay the foundation, for example “Checking and Improving Results” and “From Assistant to Agent”. By December 2027, we aim to add content specifically for this task.
Frequently Asked Questions About AI Literacy Under Article 4
Short answers to the questions HR, L&D and executives ask us most often about Article 4.
Basics
The EU AI Act is the European regulation on artificial intelligence, officially Regulation (EU) 2024/1689. It governs how AI is developed and used in the EU. As a regulation, it applies directly in all member states, without national implementing laws. It can also apply to companies outside the EU, for example when they place AI systems on the EU market or when the output of their AI systems is used in the EU.
Article 3(56) of the AI Act describes AI literacy as the skills, knowledge and understanding needed to make an informed deployment of AI systems and to be aware of the opportunities, risks and possible harm of AI. So it's about safe and productive use.
Article 4 has applied since February 2, 2025. Since August 2, 2026, national authorities have been checking compliance. On July 27, 2026, the wording was amended by Regulation (EU) 2026/1744.
Since July 27, 2026, companies have to support the development of AI literacy instead of ensuring it. The law now also states that no one has to guarantee a specific level of competence.
Yes. If you use AI tools from other providers at work, you count as a deployer and are covered. Only purely personal use is exempt.
A deployer is any company or organization that uses an AI system under its own authority. Whether you build the system yourself or buy it doesn't matter. Only purely personal use is exempt.
Yes, there is no exemption based on company size. But the amendment provides that the Commission and the member states support small and medium-sized enterprises in particular with implementation.
Yes. Besides your own staff, Article 4 also covers other persons who operate or use AI systems on your behalf, such as freelancers or service providers.
Implementation in your company
You have to take measures that develop the AI literacy of your employees. Training is the obvious way to do that. The law doesn't prescribe what it looks like.
None. There is no minimum number of hours, and no mandatory format or curriculum.
Legally, no format is prescribed. But the law requires the measures to fit a person's knowledge, experience and use of AI. The Commission confirms that different levels by role make sense.
No. The European Commission explicitly states that no certificate is needed.
It's not mandatory. But the Commission recommends keeping internal records of which trainings and measures took place. Campus makes this easier: The admin area already has everything you need, such as who has completed which courses and where each person stands. More on this under Evidence.
The AI Act doesn't set an interval. Because tools and use cases change quickly, we recommend reviewing the measures at least once a year and updating them when new tools arrive. That's exactly why Campus is a subscription: We keep the content up to date and add new topics so your employees always have the training that's relevant right now.
The AI Act obligates the company, not a specific function. In practice, the topic usually sits with people development or L&D, together with IT, legal and data protection. What matters is that someone clearly takes ownership.
Article 4 doesn't say. But in many EU countries, works councils or other employee representatives have a say when AI systems are introduced or when it comes to workplace training, for example in Germany. So involve them early.
Enforcement and penalties
The national market surveillance authorities in each EU country. They have been responsible since August 2, 2026. Some countries are still finalizing which authority takes on which role.
Article 4 has no specific fine. But authorities can order measures in individual cases, which must be proportionate. The often-cited €35 million applies to prohibited AI practices, not to Article 4. If you use AI in high-risk areas, you have additional obligations under Article 26. Fines are possible there.
Article 4 applies to any use of AI at work and requires measures on AI literacy. Article 26 applies to deployers of high-risk AI systems, for example in recruiting, credit checks or insurance. There, human oversight must lie with people who have the necessary competence, training and authority, and violations can lead to fines. For most of these systems, this applies from December 2, 2027.
Training with Agile Academy
No. No course and no provider can promise you legal certainty, including us. But a course can help show that you've taken appropriate measures. And if it's good, it moves your company forward.
Campus, our learning platform with the AI coach Henrik and the Playground, plus Live Trainings such as AI for Leaders and AI for Practitioners. The learning path differentiates by role and ranges from the basics to building your own agents.
Until December 31, 2026, Campus costs €499 per person per year, excl. VAT. On January 1, 2027, the price goes up to €999. If you book before then, you keep the current price for the entire term of your subscription. For 100 seats or more, we'll make you a custom offer.
Questions? Talk to Us
Whether it's about Article 4, Campus or the next sensible step for your company: Send us your questions or give us a call. We'll get back to you quickly.
AI Assistant
Get instant help and answers
Philip can jump in if needed.
Our Sources – What This Guide Is Based On
Legal texts
- Regulation (EU) 2026/1744, official English version in the Official Journal, published July 24, 2026, in force since July 27, 2026. Including Article 1(5) with the new wording of Article 4, and Recital 8.
- Regulation (EU) 2024/1689, the AI Act, in particular Article 3 with the definitions of deployer and AI literacy, Article 26 on the obligations of deployers of high-risk AI systems, Article 99 on penalties and Article 113 on entry into application.
Authorities and Commission
- European Commission, AI Literacy – Questions & Answers, as of July 27, 2026.
- European Commission, Living repository of AI literacy practices.
- AI Act Service Desk of the European Commission, national authorities and resources.
Studies
- McKinsey, annual survey on the state of AI in organizations, August 2026, 1,719 respondents.